Fleet Intelligence
SteamBytes — Live Fleet Overview
3 sites · Johor Bahru | Kuala Lumpur | Penang · Open-Source Platform Demo
Fleet Efficiency Trend (7D)
LIVE
Steam Output — All Sites (ton/hr)
LIVE
Active Alarms by Site
Boiler Detail
LIVE DATA
Steam Pressure Trend (bar) — Last 2 Hours
LIVE
O₂ Level vs Fuel Consumption — Last 2 Hours
CORRELATION
Digital Logbook
Operational Data Log
Replaces manual boilerman logbooks · Exportable · Auditable
Safety Logic · HAZOP
Rule Engine & Alarm Management
HAZOP safety rules ported from ThingWorx · Email/SMS notification to PIC · Tamper-proof audit log
Active Alarms — All Sites
REAL-TIME
HAZOP Event Log
AUDIT TRAIL
AI/ML Intelligence
AI/ML Capability Roadmap
3-phase roadmap from data foundation to predictive intelligence — built on the open-source platform
Phase 1 · Foundation
Data Readiness & Baseline Models
Months 1–4 · Runs in parallel with platform migration
  • Data quality audit — 16 sensor streams validated
  • TimescaleDB time-series schema deployed
  • Baseline efficiency model (linear regression)
  • Anomaly detection — statistical process control
  • Grafana ML metrics dashboard
Phase 2 · Prediction
Predictive Analytics & Optimisation
Months 5–9 · Post go-live, multi-site data available
  • Combustion efficiency predictor (XGBoost / LightGBM)
  • O₂ — fuel correlation optimiser
  • Steam demand forecasting (Prophet)
  • Predictive maintenance alerts (vibration + temp drift)
  • Python MLflow model registry deployed
Phase 3 · Intelligence
Autonomous Efficiency Optimisation
Months 10–18 · Fleet-scale multi-boiler intelligence
  • Reinforcement learning — real-time burner control advisory
  • Digital twin of boiler thermodynamics
  • Cross-fleet benchmarking & best-practice propagation
  • LLM-powered natural-language anomaly reports
  • Automated HAZOP deviation root cause analysis
Open-Source ML Stack Recommendation
Data Source
Node-RED → TimescaleDB
Retained. Time-series optimised.
Feature Store
Feast (open-source)
Reusable ML feature pipeline
ML Training
scikit-learn + XGBoost
Efficiency & anomaly models
Time-Series ML
Prophet + LSTM (PyTorch)
Demand forecasting & patterns
Model Registry
MLflow
Version control for ML models
Orchestration
Apache Airflow
Automated retraining pipelines
Visualisation
Grafana + Evidence.dev
AI insights embedded in dashboard
AI/ML Use Cases — Value Impact
UC-01
Combustion Efficiency Optimisation
Correlate O₂ level, fuel flow and steam output to find the optimal air-fuel ratio per boiler in real time.
↑ 3–5% fuel efficiency · Est. RM 85,000/year/boiler saved
UC-02
Predictive Maintenance Alerts
Detect sensor drift and component degradation 48–72h before failure — preventing unplanned shutdown.
↓ Unplanned downtime 60% · RM 120,000/incident avoided
UC-03
Steam Demand Forecasting
Predict steam demand 4–8h ahead to optimise burner load scheduling and reduce peak fuel consumption.
↓ Fuel waste 8–12% · Smoother load curves across fleet
💨 Proposal · SteamBytes RFP Response · April 2026

SteamBytes Open-Source
Platform Migration

A complete technical proposal to migrate the SteamBytes platform from PTC ThingWorx to a fully open-source, IP-sovereign stack — delivering superior dashboard intelligence, HAZOP safety logic, and a clear AI/ML roadmap for Mechmar's 50-year industrial legacy.

Prepared by: AriPrus Digicon Pvt.Ltd For: SteamBytes / Mechmar Group Contact: nazmus@steambytes.com RFP Ref: SB-RFP-2026-OSS
Why AriPrus
Our Differentiated Value
🏭
Industrial IoT Specialists
8+ years deploying open-source IIoT platforms for manufacturing, chemical processing and utilities. We understand PLCs, Modbus, Node-RED and industrial data pipelines natively — not as IT generalists.
🤖
Data Science at Industrial Scale
In-house data science team with published work on combustion efficiency ML models. We have existing model libraries for boiler O₂ optimisation and thermal anomaly detection — not starting from zero.
🔓
True IP Sovereignty
100% open-source stack — no ThingWorx, One-time license fees, no vendor lock-in.
📡
Zero-Touch Migration
Sensors, Advantech gateway and Node-RED remain 100% unchanged. We redirect the output node remotely — no site visits required for connectivity. Your 3 Malaysian sites are live in weeks, not months.
🇲🇾
Remote Support
Remote support for the pilot customer included in Phase 1. Business hours + emergency response SLA available.
⚡
Highest-Weighted: Dashboard
We take the RFP's highest-weighted criterion seriously. This demo is our answer — a production-grade dashboard that exceeds ThingWorx visualisation with fleet intelligence, interactive trend analysis and real-time HAZOP status.
Architecture
Proposed Open-Source Stack
Replacing ThingWorx entirely.
🗄
TimescaleDB
Time-series PostgreSQL — replaces ThingWorx database. Hypertables optimised for sensor data at scale.
📊
Grafana OSS
Production-grade dashboard & alerting. Connects directly to TimescaleDB. Replaces ThingWorx UI.
🔁
Apache Kafka / MQTT Broker
Receives telemetry from Node-RED. High-throughput message bus for 16 sensor streams per site.
⚙
Apache Superset
Self-service analytics and reporting layer. Exportable digital logbooks replacing manual records.
🛡
Node-RED + Nodered-contrib-rules
HAZOP safety logic implemented as Node-RED rule flows — Email/SMS PIC notification retained.
🤖
MLflow + scikit-learn + Prophet
AI/ML model training, versioning and serving. Combustion optimisation and predictive maintenance.
☁
Docker Compose on existing VPS
All services containerised. Runs on existing Mechmar cloud server. No new infrastructure cost.
🔐
Keycloak (SSO)
Role-based access for operators, managers and admins. Replaces ThingWorx user management.
Delivery Team
Team Structure
👤
Project Manager
ENGAGEMENT LEAD
PMP certified · IIoT project delivery · Client relationship · Malaysia-based · 8yr exp
👤
IoT Platform Architect
TECHNICAL LEAD
TimescaleDB · Grafana · Kafka · Node-RED · ThingWorx migration specialist · 6yr exp
👤
Full-Stack Developer
DASHBOARD & INTEGRATION
React / Vue · REST API · Node-RED flows · Industrial dashboard UI · 5yr exp
👤
Data Scientist
AI/ML LEAD
Python · scikit-learn · XGBoost · Prophet · Industrial ML · Combustion efficiency models · 4yr exp
Project Timeline
Development Milestones
16-week primary delivery · AI/ML roadmap continues in Phase 2
WEEKS 1–2
Kickoff & Architecture
NDA signed · Existing ThingWorx data schema documented · Open-source stack deployed on Mechmar VPS · TimescaleDB schema designed for 16 sensors · Node-RED connectivity plan reviewed
WEEKS 3–5
M1: Platform Live + Node-RED Integration
TimescaleDB live · Node-RED output node updated remotely to MQTT/REST · All 16 sensor streams ingested · SSL/TLS configured · First live data visible in Grafana · Deliverable: Live telemetry feed from Site 1 (wilmar)
WEEKS 6–10
M2: Full Dashboard — Fleet & Boiler Detail Views
Grafana dashboards built: Fleet Overview · Boiler Detail (Fuel/Water/Steam/Economiser) · Interactive Trends with multi-parameter overlay · Digital Data Logbook with export · Deliverable: Dashboard live and matching ThingWorx parity
WEEKS 11–13
M3: HAZOP Safety Logic + Alerting
All HAZOP rules ported from ThingWorx to Node-RED rule engine · Email/SMS gateway connected · Alarm management dashboard · PIC notification tested for all 3 sites · Deliverable: Safety logic parity confirmed by Mechmar team
WEEKS 14–16
M4: UAT + 3 Site Deployment + Handover
Sites 2 & 3 connected (KL + Penang) · Full user acceptance testing · On-site training for operators at all 3 sites · Documentation delivered · ThingWorx decommissioned · Deliverable: Full production sign-off
MONTHS 5–18
Phase 2: AI/ML Delivery (Roadmap)
Combustion efficiency ML model · Predictive maintenance alerts · Steam demand forecasting · MLflow model registry · Full Phase 2 scoped and priced separately post Phase 1 delivery
Commercial Terms
Pricing & Cost Breakdown
All prices in Malaysian Ringgit (MYR) · Subject to final scope confirmation
#Line ItemDescriptionCost (MYR)Type
1Platform ConfigurationTimescaleDB · Grafana · Kafka setup on existing VPS · SSL/TLS · Node-RED integrationRM 30,000One-time
2Dashboard DevelopmentFleet Overview · Boiler Detail · Trends · Data Logbook · Full UI exceeding ThingWorx parityRM 50,000One-time
3HAZOP Safety Logic PortAll existing HAZOP rules migrated · Email/SMS alerting · Alarm management dashboardRM 30,000One-time
43-Site Deployment & TrainingRemote Node-RED update + on-site training at wilmar, Nestle KL, IOI Penang (travel included)RM 40,000One-time
5AI/ML Phase 1 RoadmapData readiness · Baseline efficiency model · Anomaly detection · MLflow setupRM 40,000One-time
TOTAL ONE-TIME INVESTMENTRM 190,000Phase 1
6IoT Platform (Cloud/VPS)Existing Mechmar server — no new cloud needed.RM 0 / monthRecurring
7Post-Development Support12 months: 8hrs/month maintenance · bug fixes · Grafana dashboard updates · Email SLARM 4,500 / monthRecurring
8AI/ML Phase 2 (Optional)Combustion ML model · Predictive maintenance · Steam forecasting (scoped post Phase 1)RM 65,000–85,000Phase 2
Payment Terms: 30% on contract signing · 40% at M2 (Dashboard UAT sign-off) · 30% at M4 (Production go-live). 12-month warranty included. Support contract optional from Month 4.
Business Value
ROI & Business Case
Conservative estimates based on 3 sites. Mechmar fleet of 100+ boilers scales these numbers proportionally.
RM 0/mo
Platform License Cost
(vs ThingWorx)
3–5%
Fuel Efficiency Gain
via AI/ML (Phase 2)
RM 255K+
Annual Fuel Saving
3 sites · 3–5% gain
< 8 months
Full ROI Payback
Phase 1 + Phase 2
5-Year TCO: ThingWorx vs AriPrus Open-Source Stack
💨
Ready to transform SteamBytes?

This demo represents our technical answer to SteamBytes' highest-weighted criterion — dashboard excellence. We are ready to begin Week 1 activities within 5 days of contract execution.

AriPrus · info@ariprus.com · ariprus.com